Foliage penetration based on 4D LIDAR datasets

Abstract

A method for detecting terrain, through foliage, includes the steps of: receiving point cloud data in a three-dimensional (3D) space from an airborne platform, in which the point cloud data includes foliage that obscures the object; reformatting the point cloud data from the 3D space into a one-dimensional (1D) space to form a 1D signal; and decomposing the 1D signal using a wavelet transform (WT) to form a decomposed WT signal. The decomposed WT signal is reconstructed to form a low-pass filtered profile. The method classifies the low-pass filtered profile as terrain. The terrain includes a natural terrain, or a ground profile.

What is claimed: 
     
       1. A method for detecting a terrain profile using point cloud data, the method comprising the steps of:
 receiving point cloud data in a three-dimensional (3D) space from an airborne platform; 
 reformatting the point cloud data from the 3D space into a one-dimensional (1D) space to form a 1D signal; 
 decomposing the 1D signal using a wavelet transform (WT) to form a decomposed WT signal; 
 reconstructing the decomposed WT signal to form a low-pass filtered profile; and 
 classifying the low-pass filtered profile as the terrain profile; 
 wherein decomposing the 1D signal includes: 
 calculating approximation coefficients (aC) for the 1D signal, and calculating detail coefficients (dC) for the 1D signal. 
 
     
     
       2. The method of  claim 1  including the steps of:
 forming a height signal using the 1D signal; and 
 classifying a height point of the height signal as a point of an object, if the height point is above a corresponding point of the low-pass filtered profile. 
 
     
     
       3. The method of  claim 2  wherein
 the object includes a man-made object or vegetation disposed above the terrain profile. 
 
     
     
       4. The method of  claim 1  wherein
 the terrain profile includes a natural terrain profile, or a ground profile. 
 
     
     
       5. The method of  claim 1  wherein receiving the point cloud data includes:
 receiving x, y, z data from a laser detection and ranging (LADAR) system, wherein 
 x and y data are imaging data in the x and y directions of an imaging array, respectively, and 
 z data is intensity data in the z direction of the imaging array. 
 
     
     
       6. The method of  claim 5  wherein reformatting the point cloud data includes:
 dividing the imaging data into a plurality of dx strips, in which each dx strip is a narrow delta in the x direction of the imaging array, and 
 forming the 1D signal as z data in each of the plurality of dx strips. 
 
     
     
       7. The method of  claim 6  wherein
 the z data is formed by moving in the y direction as a function of each consecutive dx strip in the x direction. 
 
     
     
       8. The method of  claim 7  wherein
 the z data is formed by moving, sequentially, in an ascending order of the y direction and a descending order of the y direction as a function of each consecutive dx strip in the x direction. 
 
     
     
       9. The method of  claim 1  wherein reconstructing the decomposed WT signal includes:
 setting the detail coefficients (dC) to zero, and 
 calculating an inverse transform (W −1 ) of the WT, after setting the detail coefficients (dC) to zero, to form the low-pass filtered profile. 
 
     
     
       10. The method of  claim 9  wherein
 the decomposing step includes calculating at least three levels of aC and dC, and 
 the reconstructing step includes setting the at least three levels of dC to zero, and 
 synthesizing the at least three levels of aC to form the low-pass filtered profile. 
 
     
     
       11. The method of  claim 1  including the steps of:
 using morphological operators to further filter the terrain profile, and 
 providing the further filtered terrain profile as data to a digital terrain map (DTM). 
 
     
     
       12. The method of  claim 11  wherein
 the morphological operators include dilation and erosion. 
 
     
     
       13. The method of  claim 1  including the steps of:
 reconstructing the decomposed WT signal to form a high-pass filtered profile; and 
 classifying the high-pass filtered profile as discontinuities in the terrain profile; 
 wherein the discontinuities denote edges of man-made structures. 
 
     
     
       14. The method of  claim 13  wherein decomposing the 1D signal includes:
 calculating approximation coefficients (aC) for the 1D signal, and 
 calculating detail coefficients (dC) for the 1D signal; and 
 reconstructing the decomposed WT signal includes: 
 setting the approximation coefficients (aC) to zero, and 
 calculating an inverse transform (W −1 ) of the WT, after setting the 
 approximation coefficients (aC) to zero, to form the high-pass filtered profile. 
 
     
     
       15. The method of  claim 1  wherein
 the WT includes a discrete WT, a continuous WT, a stationary WT and a multilevel wavelet decomposition (wavedec). 
 
     
     
       16. A method for detecting a terrain profile, through foliage, using point cloud data, the method comprising the steps of:
 receiving point cloud data in a three-dimensional (3D) space from an airborne platform; 
 reformatting the point cloud data from the 3D space into a two-dimensional (2D) space to form a 2D signal; 
 decomposing the 2D signal using a wavelet transform (WT) to form a decomposed WT signal; 
 reconstructing the decomposed WT signal to form a low-pass filtered profile; and 
 classifying the low-pass filtered profile as the terrain profile; 
 including the steps of: 
 attenuating high frequency components of the point cloud data in the 3D space to form a filtered height signal; 
 defining the low-pass filtered profile as a ground reference signal; and 
 calculating a distance between a point on the filtered height signal and a corresponding point on the ground reference signal to determine whether the point on the filtered height signal belongs to a ground class or an object class. 
 
     
     
       17. The method of  claim 16  wherein the object class includes man-made objects or vegetation, and
 the ground class includes natural terrain. 
 
     
     
       18. The method of  claim 16  wherein a a predetermined threshold value is used in determining whether a point belongs to the ground class or the object class.